Date: Fri, 28 Jan 2005 10:42:18 -0600
Reply-To: Robin High <robinh@UNLSERVE.UNL.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Robin High <robinh@UNLSERVE.UNL.EDU>
Subject: Re: Stat Question--number continuous nominal categorical etc.
Content-Type: TEXT/PLAIN; charset=US-ASCII
> Let's take age from 1 to 100. Clearly this variable is a
> number/continuous variable. ^^^^^^^
..if you measure to the nearest month, day, hour, minute, second,.... it
is "clear", though even when rounded to the nearest year, if you have a
range of values for age, continuous data analysis techniqes are usually
reasonably good approximations. Age clustering around a few numbers or
approximations can cause difficulties.
> Now let's bucket the age variable into, say, 15 buckets or however
> many buckets you wish. How would you classify that age variable now?
> Would you say it is a number/ordinal? No, bec. there is no implied
> order or significance in the categories. Would you say it is
> number/continuous? I don't know. How about number/nominal? Would you
> classiffy the bucketed age as number/nominal?
How is it not ordinal? Categorizing age into 'bins' (or buckets) is a
common approach, though generally you are adding measurement error by
doing so (which biases results).
> In my situation I have MSAs. Each MSA (bucket) has the same $amount
> many times. One MSA may have $934456 20 thousand times. The next MSAS
> may have the $3678 5 thousand times. The next MSA may have .... you
> see the picture. Is this variable number/continuous? Is it
.. You seem to have large numbers with very precise measurements -- its
hard to believe any process would generate exactly 934456 thousands of
time. How many 'discrete' values like this do you have and what analysis
techniques or types of hypotheses are you considering? Treating it as an
ordinal measurement would be better than continuous (in my opinion) though
I don't have enough background on what you are trying to do with this data
to make that a recommendation.
Univ. of Oregon